Nothing Special   »   [go: up one dir, main page]

US6996266B2 - Method and apparatus for generating three-dimensional data of an object by selecting the method for detecting corresponding points that is suitable to the object - Google Patents

Method and apparatus for generating three-dimensional data of an object by selecting the method for detecting corresponding points that is suitable to the object Download PDF

Info

Publication number
US6996266B2
US6996266B2 US09/925,742 US92574201A US6996266B2 US 6996266 B2 US6996266 B2 US 6996266B2 US 92574201 A US92574201 A US 92574201A US 6996266 B2 US6996266 B2 US 6996266B2
Authority
US
United States
Prior art keywords
detector
image data
dimensional data
corresponding points
images
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Lifetime, expires
Application number
US09/925,742
Other versions
US20020018062A1 (en
Inventor
Takayuki Hamaguchi
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Minolta Co Ltd
Original Assignee
Minolta Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Minolta Co Ltd filed Critical Minolta Co Ltd
Assigned to MINOLTA CO. LTD. reassignment MINOLTA CO. LTD. ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: HAMAGUCHI, TAKAYUKI
Publication of US20020018062A1 publication Critical patent/US20020018062A1/en
Application granted granted Critical
Publication of US6996266B2 publication Critical patent/US6996266B2/en
Adjusted expiration legal-status Critical
Expired - Lifetime legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/50—Depth or shape recovery
    • G06T7/55—Depth or shape recovery from multiple images
    • G06T7/593—Depth or shape recovery from multiple images from stereo images
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
    • H04N13/20—Image signal generators
    • H04N13/204—Image signal generators using stereoscopic image cameras
    • H04N13/239—Image signal generators using stereoscopic image cameras using two 2D image sensors having a relative position equal to or related to the interocular distance
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
    • H04N13/20—Image signal generators
    • H04N13/204—Image signal generators using stereoscopic image cameras
    • H04N13/243—Image signal generators using stereoscopic image cameras using three or more 2D image sensors
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G06T2207/10012—Stereo images
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
    • H04N13/10—Processing, recording or transmission of stereoscopic or multi-view image signals
    • H04N13/189—Recording image signals; Reproducing recorded image signals
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
    • H04N2013/0074—Stereoscopic image analysis
    • H04N2013/0081—Depth or disparity estimation from stereoscopic image signals

Definitions

  • the present invention relates to a method and an apparatus for generating three-dimensional data by detecting corresponding points of two two-dimensional images whose contents are similar to each other such as two two-dimensional images captured stereographically with respect to an object using a gradient method, a correlation method or the like.
  • the corresponding points are detected by setting an image as a standard image and finding an area similar to the standard image from another image. For example, the corresponding points are detected in such a manner by extracting a small area of the standard image as a standard window and detecting, in another image, a window having luminance of points similar to those of points of the standard window.
  • the correlation method has other applications such as a method for detecting the corresponding points by a correlation of luminance on an epipolar line, a method wherein the size of a window is changed and a method wherein a correlation is detected by binarization of an image.
  • the gradient method Another known method for detecting the corresponding points is the gradient method.
  • the corresponding points are detected from luminance gradient of an area adjacent to a point whose coordinate is in common to plural images similar to each other.
  • the correlation method is superior in detecting corresponding points of an object whose luminance or color gradation is widely varying on its surface.
  • any of the above methods are not fully satisfactory for the detection of the corresponding points of an object that is poor in luminance change or an object that is low in contrast. Therefore, the correlation method is not suitable for generating three-dimensional data of a face of a human or the like.
  • the gradient method requires shorter calculation time as compared with the correlation method and is superior in detecting the corresponding points in the case of low luminance and low contrast. Further, the gradient method can detect the corresponding points correctly by using subpixels if the object is in an ideal state. However, the gradient method is not competent to differences in luminance or color gradation or CCD noise and has a drawback that an erroneous detection of the corresponding points occurs when there is an occlusion. Therefore, the gradient method is not suitable for generating three-dimensional data of products having a spherical or cylindrical shape, products made from a metal or glass and so on.
  • An object of the present invention is to generate highly reliable three-dimensional data of any types of objects such as a human face, an industrial product and so on.
  • the three-dimensional data generating apparatus comprises: an obtaining portion for obtaining plural image data representing images obtained by photographing an object from different viewpoints, the plural image data each representing respective one of the images; a first detector for detecting a corresponding point in each of the images by the gradient method; a second detector for detecting a corresponding point in each of the images by the correlation method; a selector for selecting either one of the first detector or the second detector; and a generator for generating three-dimensional image of the object based on the corresponding point detect by the selected detector.
  • the apparatus may comprise an object judging portion for judging a type of an object, and the selector may select either one of the first detector or the second detector based on a judgment result obtained by the object judging portion.
  • the obtaining portion may be comprised in the apparatus separately from other portions or integrally with other portions.
  • FIG. 1 shows an appearance of a three-dimensional image generating apparatus according to a first embodiment of the present invention.
  • FIG. 2 is a block diagram showing a structure of the three-dimensional data generating apparatus according to the first embodiment.
  • FIG. 3 is an example of two images.
  • FIG. 4 is a graph showing an expression of a luminance gradient.
  • FIG. 5 is an enlarged view of windows WA and WB.
  • FIG. 6 shows a noted pixels GA and GB as well as pixels in the vicinity of the noted pixels.
  • FIG. 7 illustrates a method wherein a multiple resolution image is used by way of an example.
  • FIG. 8 shows a concept of detection using an epipolar line EL.
  • FIG. 9 illustrates a relationship between a correlation value and difference.
  • FIG. 10 is a flowchart showing a flow of the generation of three-dimensional data according to the first embodiment.
  • FIG. 11 is a block diagram showing a three-dimensional data generating apparatus according to a second embodiment of the present invention.
  • FIG. 12 is a flowchart showing processing for generating three-dimensional data according to the second embodiment.
  • FIG. 13 shows an appearance of a three-dimensional data generating apparatus 1 C according to the present invention.
  • FIG. 14 is a block diagram showing constitution of the three-dimensional data generating apparatus 1 C according to the third embodiment.
  • FIG. 15 illustrates the method of obtaining reliability of the detection results by way of an example.
  • FIG. 16 is a flowchart showing a flow of processing for obtaining the reliability of the detection results.
  • FIG. 17 is a flowchart showing a flow of processing for generating the three-dimensional data according to a third embodiment.
  • FIG. 18 shows a three-dimensional data generating apparatus as a modification.
  • colors of an object are detected to select a detector for generating three-dimensional data.
  • FIG. 1 shows an appearance of the three-dimensional data generating apparatus 1 according to the first embodiment.
  • FIG. 2 is a block diagram showing a structure of the three-dimensional data generating apparatus 1 according to the first embodiment.
  • the three-dimensional data generating apparatus 1 comprises two image input portions 21 , processor 3 A and a detector appointing portion 4 .
  • the image input portions 21 each comprises an optical system 12 p , an imaging element 12 q , a camera 12 including a control circuit not shown and so on.
  • the image input portions 21 serves to generate image data.
  • the optical system 12 p comprises various lenses, a diaphragm and so on and serves to form an image on the imaging element 12 q by receiving light from an object Q.
  • the imaging element 12 q converts the image of the object Q into an electric signal.
  • the imaging element 12 q there may be used a CCD type solid-state imaging element and the like.
  • the image signal generated by the imaging element 12 q is converted into a digital signal by an A/D converter so as to be image data.
  • the detector appointing portion 4 is a sliding switch for appointing either one of detectors to be described later in this specification in accordance with manual operation of a user. This sliding switch can be switched among 3 modes of “portrait”, “auto” and “landscape”.
  • the processor 3 A comprises various devices such as a display 11 , an input device, a CPU, a RAM, a ROM, a control circuit, software such as a control program and the like.
  • a display 11 an input device
  • a CPU central processing unit
  • a RAM random access memory
  • a ROM read-only memory
  • a control circuit software such as a control program and the like.
  • the image data memories 31 each memorize the image data generated by the image input portion.
  • the object judging portion 32 S judges a type of the object Q by detecting colors of the object Q and examining whether or not the object Q comprises a predetermined color in a predetermined amount.
  • Amounts of colors to be comprised in the object Q are decided by determining an average of color information of pixels comprised in the image data obtained by the image input portions 21 . Characteristics of the object Q appear in the vicinity of center of an image and, therefore, only the amounts of colors in the vicinity of the center of the object Q may be calculated for the judgment of the type of the object Q. In order to judge a type of the object Q, the image data may be low in resolution since they are used only for the detection of colors and the calculation of the amounts of colors.
  • r(R), g(G), b(B) each is a function representing an amount of each of three primitive colors of light: red, green and blue.
  • R, G and B that are parameters for the function.
  • the functions r(R), g(G) and b(B) are increased in an amount of red, green and blue, respectively, in proportion to the parameters R, G and B.
  • c(C) an area in the image data such as c(C) of each pixels in the vicinity of the center of the image corresponds to a specific color. It is judged that the object is a human when c(C) is a color of seasan in the present embodiment since the present embodiment is designed for a xanthoderm.
  • Conditions under which c(C) be seasan are R ⁇ 250, G ⁇ 240 and B ⁇ 150. These values or areas specified by the values can be varied depending on the photographing conditions when so required.
  • other colors may be set as c(C) instead of seasan depending on the race of human.
  • Plural colors may be set as c(C) on the condition of logical OR so that plural races may be recognized as human.
  • the detector selecting portion 33 S selects, based on the result obtained from the judgment by the object judging portion 32 S, either one of the first corresponding point detector 34 which employs the gradient method or the second corresponding point detector 35 which employs the correlation method for the detection of corresponding points. For example, in the case where the object Q is a human, the first corresponding point detector 34 is selected; in the case where the object Q is not a human, the second corresponding point detector 35 is selected.
  • the object judging portion 32 S does not perform the judgment processing.
  • the detector selecting portion 33 S operates in such a manner that the first corresponding point detector 34 is selected if the “portrait” is selected and the second corresponding point detector 35 is selected if the “landscape” is selected.
  • the first corresponding point detector 34 detects corresponding points of plural image data DT 2 by the gradient method.
  • the second corresponding point detector 35 detects corresponding points of plural image data DT 2 by the correlation method. Descriptions on the detectors will be given later in this specification.
  • the three-dimensional data generator 36 generates three-dimensional data DT 3 based on the detection result obtained by either one of the first corresponding point detector 34 or the second corresponding point detector 35 . More specifically, the three-dimensional data DT 3 are generated by reproducing three-dimensional image of the object based on the corresponding points of the plural image data DT 2 .
  • the three-dimensional data memory 37 memorizes the three-dimensional data DT 3 .
  • FIG. 3 is an example of plural image IM 2 ;
  • FIG. 4 is a graph showing an expression of luminance gradient;
  • FIG. 5 is an enlarged view of windows WA and WB;
  • FIG. 6 shows noted pixels GA and GB as well as pixels in the vicinity of the noted pixels; and
  • FIG. 7 illustrates an example of the method wherein a multiple resolution image is used.
  • Images IM 2 a and IM 2 b are images that are represented by each of the plural image data DT 2 a and DT 2 b .
  • the image data DT 2 a and DT 2 b are data generated by separate image input portions 21 .
  • the image IM 2 b represents a standard image and the image IM 2 a represents a reference image.
  • the images IM 2 a and IM 2 b respectively include the object image QA and QB of the object Q.
  • the windows WA and WB are set in each of the images IM 2 a and IM 2 b .
  • the window WB is a standard window set in the standard image
  • the window WA is a reference window set in the reference image.
  • the windows WA and WB are at a position that includes right eye of the object.
  • each of the windows WA and WB is comprised of 9 ⁇ 9 pixels.
  • Purpose for detecting the corresponding points is to find on the reference image points corresponding to all of points on the object image QB of the object Q in the standard image.
  • Vector connecting the corresponding points when the reference image and the standard image are superimposed is referred to as a “corresponding vector” in this specification, and the detection of corresponding points using the gradient method means the detection of corresponding vector (u, v) with respect to points on an image.
  • the corresponding vector has a point of the standard image as its base point and has a point on the reference image as its end point.
  • the object images QA and QB of FIG. 3 are prepared by photographing almost the same portion of the object Q, it is possible to assume that the luminance gradient of a point on a coordinate of the object image QA is the same as that of a point on the same coordinate of the object image QB.
  • the expression (2) contains two unknown letters u and v while containing only one equation, it can represent nothing more than that corresponding vectors are on a linear line as shown in FIG. 4 . Therefore, another restraining condition is required between u and v in order to determine the corresponding vectors. Thus, there is made such an assumption that u and v are common to all the pixels in a certain area, for example, in an area of 3 ⁇ 3 pixels. Solution for u and v with respect to all the pixels are obtainable by solving, by the least square method, a simultaneous equation that is made by adding the restraining condition to the expression (2). It is possible to reduce influence of quantization error in imaging or noise of imaging element by using a wider area such as 5 ⁇ 5 pixels and 6 ⁇ 6 pixels.
  • the corresponding vectors with respect to all the pixels are the same, there can be prepared 81 types of the expression (2).
  • the corresponding vectors are obtainable by solving the expressions by using the least square method.
  • Ex, Ey and Et of the expression (2) with respect to the pixels GA and GB can be represented by the following expressions (4) to (6).
  • Ex ⁇ ([ GA 1 ] ⁇ [ GA 2 ])+([ GB 1 ] ⁇ [GB 2 ]) ⁇ /2 (4)
  • Ey ⁇ ([ GA 3 ] ⁇ [ GA 4 ])+([ GB 3 ] ⁇ [ GB 4 ]) ⁇ /2 (5)
  • Et [GB] ⁇ [GA] (6)
  • the reference characters such as [GA 1 ] respectively represent luminance values of the pixels shown in FIG. 6 .
  • [GA 1 ] represents the luminance value of the pixel GA 1 shown in FIG. 6 .
  • the multi resolution image is utilized in order to detect the corresponding points precisely.
  • the method using the multi resolution image it self is known in the art.
  • the multiple resolution image FM comprises an original image F and a plurality of types of images F 1 , F 2 and F 3 obtained by so compressing the original image F as to be (1 ⁇ 2) 2 , (1 ⁇ 4) 2 , (1 ⁇ 8) 2 or the like that are low in resolution.
  • An image that is relatively low in resolution typically has smaller number of pixels than that of an image that is high in resolution. Therefore, pyramid shape is formed by arranging an image that is the highest in resolution at the bottom and other images in the descending order of resolution.
  • the multiple resolution image FM forms an image pyramid.
  • the multiple resolution image FM By the use of the multiple resolution image FM, two positions distant to each other between images IM 2 a and IM 2 b when the resolution is high, i.e., when there is a large number of pixels, approaches to each other when there is a small number of pixels. Accordingly, it is possible to widen an area wherein the corresponding points are precisely obtained by the gradient method as a result of transmitting the difference of images from a low resolution to a high resolution.
  • the corresponding vector is obtained with respect to the lowest resolution image to be set as an initial value. Then, another corresponding vector is obtained with respect to the second lowest resolution image.
  • FIG. 8 shows a concept of detection using an epipolar line EL.
  • FIG. 9 illustrates a relationship between a correlation value and difference.
  • the correlation method is a method for detecting windows that are high in correlation and is useful for calculating correlation between windows of two images.
  • a window that is the lowest in the correlation value is detected first.
  • the correlation value is obtainable by adding differences between corresponding points of the windows.
  • the corresponding windows are on the epipolar line and, therefore, the detection is conducted solely with respect to the epipolar line.
  • the window Wa is moved in the image IM 2 a from left upper end to right bottom end, and a window Wb corresponding the window Wa is detected from the image IM 2 b at each of the positions.
  • the window Wb in the image IM 2 b is shifted by one pixel from the same coordinate as that of the image IM 2 a .
  • the correlation value of the windows Wa and Wb is calculated to detect a position which is the lowest in the correlation value.
  • the difference obtained hereby is then obtained with respect to all the points moved in the image IM 2 a to form a map of differences.
  • the differences correspond to parallax in the case of the stereographical view and thus called “parallax map”.
  • the difference is used to enter the horizontal axis and the correlation value is used to enter the vertical axis.
  • the correlation value is the lowest when the difference is detected with respect to seven pixels. Accordingly, it is understood from the graph that a window that is seven pixels off from the noted window is the most correlative to the noted window. Parallax obtained by above calculation is seven pixels.
  • parallax is obtainable with respect to pixels in the correlation method.
  • the parallax does not always correspond to number of pixels.
  • the graph of solid line is interpolated.
  • One of the methods for the interpolation is a method wherein four points close to the lowest correlation value are noted and a point of intersection of lines formed by connecting each two points of the four points is set as the lowest correlation value.
  • the parallax corresponds to 7.3 pixels as shown by the broken line in FIG. 9 .
  • FIG. 10 is the flowchart showing a flow of the generation of three-dimensional data DT 3 according to the first embodiment.
  • two sets of image data DT 2 are obtained by the two image input portions 21 (# 11 ).
  • the camera 12 used for photographing the object Q is subjected to calibration in advance of the photographing in order to grasp camera parameters such as projection matrix and lens aberration.
  • the object is shoot in succession by monocular vision using one camera.
  • Amounts of colors of the object Q are detected based on the obtained image data DT 2 (# 12 ) to judge whether or not the object is a human (# 13 ).
  • a corresponding point of the image data DT 2 a and DT 2 b is detected by the first corresponding point detector 34 , i.e. by the gradient method (# 14 ). If it is not judged that the object Q is a human, the corresponding point of the image data DT 2 a and DT 2 b is detected by the second corresponding point detector 35 , i.e. by the correlation method (# 15 ).
  • three-dimensional data DT 3 are generated based on either of the detection results of the step 14 or the step 15 (# 16 ).
  • the object is a human based on any colors. Further, if the color r(R) solely is extracted from pixels of the object Q. characteristics of eyes, nose, forehead or cheeks are represented by red color gradation. It is possible to judge whether or not the object Q is a human by the red color gradation. The judgment may be made based on other various information such as infrared rays that radiates from the object by other various methods.
  • the selection of detector for generating the three-dimensional data is performed by detecting colors of the object.
  • the selection is performed by calculating luminance gradient in the second embodiment.
  • FIG. 11 is a block diagram showing a three-dimensional data generating device 1 B according to the second embodiment of the present invention.
  • the three-dimensional data generating apparatus 1 B comprises two image input portions 21 and a processor 3 B.
  • the image input devices are the same as those of the first embodiment.
  • the processor 3 B of the second embodiment comprises a luminance gradient calculator 32 T and a detector selecting portion 33 T unlike the processor 3 A of the first embodiment.
  • the two image data memories 31 , the first corresponding detector 34 , the second corresponding point detector 35 , the three-dimensional data generator 36 , the three-dimensional data memory 37 and the image display 38 are the same as those of the first embodiment.
  • Operation processing is carried out in the luminance gradient calculator 32 T only when the “auto” is appointed by the detector appointing portion 4 , and the detector selecting portion 33 T executes the selection based on the calculation results in the same manner as the first embodiment.
  • the first corresponding point detector 34 or the second corresponding point detector 35 is selected based on the appointment in the same manner as the first embodiment.
  • the luminance gradient calculator 32 T calculates a luminance gradient of an object based on image data DT 2 .
  • a luminance gradient Ex in the horizontal direction is obtainable by the expression (4), and a luminance gradient in the vertical direction Ey is obtainable by the expression (5).
  • the luminance gradients Ex and Ey are calculated with respect to a several points of the image data DT 2 , and the luminance gradient of the object is calculated by, for example, obtaining a mean luminance gradient of the points.
  • Two sets of image data DT 2 used in the expressions (4) and (5) are obtained by photographing almost the same part and, therefore, it is possible to use only one of them for the calculation.
  • the detector selecting portion 33 T selects either one of the first corresponding pint detector 34 or the second corresponding point detector 35 based on the judgment result obtained by the luminance gradient calculator 32 T.
  • the luminance gradient calculator 32 T performs the selection in such a manner that the first corresponding point detector 34 is selected when the luminance gradient is lower than a predetermined value while the second corresponding point detector 35 is selected when the luminance gradient is higher than the predetermined value.
  • FIG. 12 is the flowchart showing the processing for generating three-dimensional data according to the second embodiment.
  • two sets of image data are obtained by two image input portions 21 in the same manner as the first embodiment (# 21 ).
  • a luminance gradient of an object is calculated based on the obtained image data (# 22 ), and it is judged whether or not a result of the calculation is higher than a predetermined value (# 23 ).
  • a corresponding point of the two sets of image data is detected by using the second detector— 35 , i.e. by the correlation method (# 24 ).
  • the corresponding point of the two sets of image data is detected by the first detector— 34 , i.e. by the gradient method (# 25 ).
  • generation of three-dimensional data is performed by using either one of the detectors.
  • both of the detectors are used for the detection of corresponding point and the generation of three-dimensional data is performed by using detection result of higher reliability.
  • FIG. 13 shows an appearance of a three-dimensional data generating apparatus 1 C according to the third embodiment.
  • FIG. 14 is a block diagram showing a structure of the three-dimensional data generating apparatus 1 C according to the third embodiment.
  • the three-dimensional image generating apparatus 1 C comprises three image input portions 21 and a processor 3 C.
  • the image input portions operates in the same manner as those of the first embodiment.
  • the processor 3 C is comprised of the same components as those of the first embodiment; however, the processor 3 C is provided with three image data memories 31 , a reliability operation portion 32 U, a reliability judging portion 33 U and a three-dimensional data generator 36 U. Functions of the first corresponding point detector 34 , the second corresponding point detector 35 , the three-dimensional data memory 37 , the image display 38 and the like are realized in the same manner as the first embodiment.
  • the reliability judging portion 33 U compares a detection result obtained by the first corresponding point detector with a detection result obtained by the second corresponding point detector to judge which detection result is higher in reliability.
  • the three-dimensional data generator 36 U performs three-dimensional reproduction based on the detection result higher in reliability based on the judgment result of the reliability judging portion 33 U.
  • FIG. 15 illustrates an example of the method of obtaining reliability of the detection results.
  • the image data DT 2 e , DT 2 f and DT 2 g are respectively generated by different image input portions.
  • the point Pt 1 is an arbitrary point included in the image data DT 2 e .
  • the points Pt 2 and Pt 3 respectively are obtained by the first corresponding point detector and the second corresponding point detector as points on the image data DT 2 f and DT 2 g corresponding to the point Pt 1 .
  • the point Pt 12 is a point in a three-dimensional space obtained based on the points Ptl and Pt 2 .
  • the point Pt 3 ′ is a point obtained by projecting the point Pt 12 on the image data DT 2 g.
  • FIG. 16 is a flowchart showing a flow of processing for obtaining the reliability of detection results.
  • FIG. 16 there are obtained the points Pt 2 and Pt 3 that are on the image data DT 2 f and DT 2 g , respectively, and correspond to the point Pt 1 on the image data DT 2 e (# 91 ).
  • the point Pt 12 on a three-dimensional coordinate is calculated based on the points Pt 1 and Pt 2 (# 92 ).
  • the point Pt 12 is then projected onto the image data DT 2 g to obtain Pt 3 ′ (# 93 ).
  • the points Pt 3 and Pt 3 ′ are logically the same; however, in reality, the points Pt 3 and Pt 3 ′ are different from each other due to calibration error or characteristics of the camera 12 .
  • the difference between the point Pt 3 and the point Pt 3 ′ is determined by comparison thereby to obtain reliability of the detection result (# 94 ). Smaller the difference is, higher is the reliability.
  • FIG. 17 is the flowchart showing a flow of the processing for generation of the three-dimensional data according to the third embodiment.
  • three sets of image data are obtained by the three image input portions 21 (# 31 ).
  • Detection of corresponding points are performed by the first corresponding point detector 34 and the second corresponding point detector 35 (# 32 ), and reliability of the detection results are calculated (# 33 ).
  • the three-dimensional data are generated based on the selected detection result (# 37 ).
  • the reliability of the detection result obtained by the first corresponding point detector 34 is compared with the reliability of the detection result obtained by the second corresponding point detector 35 and the three-dimensional data are generated based on the detection result higher in the reliability, it is possible to generate highly reliable three-dimensional data according to the third embodiment.
  • the comparison of the reliability of the detection results is performed with respect to whole parts of the detection results in this embodiment; however, the comparison may be performed with dividing the detection results into partial areas.
  • reliability of detection result of the first corresponding point detector 34 and reliability of detection result of the second corresponding point detector 35 are obtained with respect to a plurality of partial areas, and the reliabilities are compared to each other with respect to each of the partial areas to select the higher reliability with respect to each partial areas. Then, detection results higher in the reliability are so synthesized as to complete three-dimensional reproduction, thereby to generate the three-dimensional data.
  • parts corresponding to a background may be deleted from the image data in advance of the generation of three-dimensional data. Processing of the deletion of background may be carried out by various known methods.
  • the deletion of background enables to eliminate redundant data and thus to reduce erroneous corresponding points.
  • FIG. 18 is a block diagram showing the modification of three-dimensional data generating apparatus.
  • a computer such as a personal computer is used for the processor 3 .
  • a multiple-lens camera 2 has a plurality of image input portions 21 and output portions 22 A, 22 B and 22 C for outputting image signals obtained by the image input portions.
  • Reference numeral 22 A denotes a connection port for transferring the image signals to an external device via interface such as USB; reference numeral 22 B denotes an output port for transferring the image signals to an external device via wireless communication such as infrared communication and reference numeral 22 C denotes a driver for recording the image signals to a portable recording medium such as a memory card.
  • a computer 3 is provided with input portions 30 A, 30 B and 30 C that correspond respectively to the output portions 22 A, 22 B and 22 C.
  • the input portion 30 A is a connection port for receiving signals from an external device via interface; the input portion 30 B is an input port for receiving signals from an eternal device via wireless communication and the input portion 30 C is a driver for reading data memorized in a portable recording medium.
  • An image data memory 31 serves to receive the image signals input into the input portions and to store the same.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)
  • Length Measuring Devices By Optical Means (AREA)
  • Image Analysis (AREA)

Abstract

A three-dimensional data generating apparatus of the present invention comprises an obtaining portion for obtaining plural image data representing images obtained by photographing an object from different viewpoints, the plural image data each representing respective one of the images; a first detector for detecting corresponding points of the images by a gradient method; a second detector for detecting corresponding points of the images by a correlation method; a selector for selecting either one of the first detector or the second detector; and a generator for generating three-dimensional image of the object based on the corresponding points detect by the selected detector. The present invention is capable of generating highly reliable three-dimensional data irrelevant to a type of an object.

Description

This applicatoin is based on Japanese Patent Application No. 2000-243860 filed on Aug. 11, 2000, the contents of which are hereby incorporated by reference.
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a method and an apparatus for generating three-dimensional data by detecting corresponding points of two two-dimensional images whose contents are similar to each other such as two two-dimensional images captured stereographically with respect to an object using a gradient method, a correlation method or the like.
2. Description of the Prior Art
Conventionally, there are known techniques for generating three-dimensional data from plural two-dimensional image data (a two-dimensional image is hereinafter referred to as an “image” and two-dimensional image data are referred to as “image data” in the present specification) as of an object captured by, for example, using a plurality of cameras. In generating the three-dimensional data, it is necessary to detect to which point of an image an arbitrary point of another image corresponds to, i.e. to detect corresponding points of the images. The correlation method is the most popular method for detecting such corresponding points.
By the correlation method, the corresponding points are detected by setting an image as a standard image and finding an area similar to the standard image from another image. For example, the corresponding points are detected in such a manner by extracting a small area of the standard image as a standard window and detecting, in another image, a window having luminance of points similar to those of points of the standard window. The correlation method has other applications such as a method for detecting the corresponding points by a correlation of luminance on an epipolar line, a method wherein the size of a window is changed and a method wherein a correlation is detected by binarization of an image.
Another known method for detecting the corresponding points is the gradient method. In the gradient method, the corresponding points are detected from luminance gradient of an area adjacent to a point whose coordinate is in common to plural images similar to each other.
The correlation method is superior in detecting corresponding points of an object whose luminance or color gradation is widely varying on its surface. However, any of the above methods are not fully satisfactory for the detection of the corresponding points of an object that is poor in luminance change or an object that is low in contrast. Therefore, the correlation method is not suitable for generating three-dimensional data of a face of a human or the like.
In turn, the gradient method requires shorter calculation time as compared with the correlation method and is superior in detecting the corresponding points in the case of low luminance and low contrast. Further, the gradient method can detect the corresponding points correctly by using subpixels if the object is in an ideal state. However, the gradient method is not competent to differences in luminance or color gradation or CCD noise and has a drawback that an erroneous detection of the corresponding points occurs when there is an occlusion. Therefore, the gradient method is not suitable for generating three-dimensional data of products having a spherical or cylindrical shape, products made from a metal or glass and so on.
SUMMARY OF THE INVENTION
An object of the present invention is to generate highly reliable three-dimensional data of any types of objects such as a human face, an industrial product and so on.
The three-dimensional data generating apparatus according to the present invention comprises: an obtaining portion for obtaining plural image data representing images obtained by photographing an object from different viewpoints, the plural image data each representing respective one of the images; a first detector for detecting a corresponding point in each of the images by the gradient method; a second detector for detecting a corresponding point in each of the images by the correlation method; a selector for selecting either one of the first detector or the second detector; and a generator for generating three-dimensional image of the object based on the corresponding point detect by the selected detector.
Preferably, the apparatus may comprise an object judging portion for judging a type of an object, and the selector may select either one of the first detector or the second detector based on a judgment result obtained by the object judging portion.
The obtaining portion may be comprised in the apparatus separately from other portions or integrally with other portions.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 shows an appearance of a three-dimensional image generating apparatus according to a first embodiment of the present invention.
FIG. 2 is a block diagram showing a structure of the three-dimensional data generating apparatus according to the first embodiment.
FIG. 3 is an example of two images.
FIG. 4 is a graph showing an expression of a luminance gradient.
FIG. 5 is an enlarged view of windows WA and WB.
FIG. 6 shows a noted pixels GA and GB as well as pixels in the vicinity of the noted pixels.
FIG. 7 illustrates a method wherein a multiple resolution image is used by way of an example.
FIG. 8 shows a concept of detection using an epipolar line EL.
FIG. 9 illustrates a relationship between a correlation value and difference.
FIG. 10 is a flowchart showing a flow of the generation of three-dimensional data according to the first embodiment.
FIG. 11 is a block diagram showing a three-dimensional data generating apparatus according to a second embodiment of the present invention.
FIG. 12 is a flowchart showing processing for generating three-dimensional data according to the second embodiment.
FIG. 13 shows an appearance of a three-dimensional data generating apparatus 1C according to the present invention.
FIG. 14 is a block diagram showing constitution of the three-dimensional data generating apparatus 1C according to the third embodiment.
FIG. 15 illustrates the method of obtaining reliability of the detection results by way of an example.
FIG. 16 is a flowchart showing a flow of processing for obtaining the reliability of the detection results.
FIG. 17 is a flowchart showing a flow of processing for generating the three-dimensional data according to a third embodiment.
FIG. 18 shows a three-dimensional data generating apparatus as a modification.
 DESCRIPTION OF THE PREFERRED EMBODIMENTS
(First Embodiment)
In the first embodiment, colors of an object are detected to select a detector for generating three-dimensional data.
FIG. 1 shows an appearance of the three-dimensional data generating apparatus 1 according to the first embodiment. FIG. 2 is a block diagram showing a structure of the three-dimensional data generating apparatus 1 according to the first embodiment.
As shown in FIGS. 1 and 2, the three-dimensional data generating apparatus 1 comprises two image input portions 21, processor 3A and a detector appointing portion 4.
Referring to FIG. 2, the image input portions 21 each comprises an optical system 12 p, an imaging element 12 q, a camera 12 including a control circuit not shown and so on. The image input portions 21 serves to generate image data. The optical system 12 p comprises various lenses, a diaphragm and so on and serves to form an image on the imaging element 12 q by receiving light from an object Q. The imaging element 12 q converts the image of the object Q into an electric signal. As the imaging element 12 q, there may be used a CCD type solid-state imaging element and the like. The image signal generated by the imaging element 12 q is converted into a digital signal by an A/D converter so as to be image data.
The detector appointing portion 4 is a sliding switch for appointing either one of detectors to be described later in this specification in accordance with manual operation of a user. This sliding switch can be switched among 3 modes of “portrait”, “auto” and “landscape”.
The processor 3A comprises various devices such as a display 11, an input device, a CPU, a RAM, a ROM, a control circuit, software such as a control program and the like. By above devices and so on, there can be realized functions of two image data memories 31, an object judging portion 32S, a detector selecting portion 33S, a first corresponding point detector 34, a second corresponding point detector 35, a three-dimensional data generator 36, a three-dimensional data memory 37, an image display 38 and so on.
The image data memories 31 each memorize the image data generated by the image input portion.
The object judging portion 32S judges a type of the object Q by detecting colors of the object Q and examining whether or not the object Q comprises a predetermined color in a predetermined amount.
Amounts of colors to be comprised in the object Q are decided by determining an average of color information of pixels comprised in the image data obtained by the image input portions 21. Characteristics of the object Q appear in the vicinity of center of an image and, therefore, only the amounts of colors in the vicinity of the center of the object Q may be calculated for the judgment of the type of the object Q. In order to judge a type of the object Q, the image data may be low in resolution since they are used only for the detection of colors and the calculation of the amounts of colors. The color information is typically obtained by a color equation c(C) of the following expression (1):
c(C)=r(R)+g(G)+b(B)  (1):
Here, r(R), g(G), b(B) each is a function representing an amount of each of three primitive colors of light: red, green and blue. When r(R), g(G) and b(B) each has 8 bits, an integer from 0 to 255 is assigned to each of R, G and B that are parameters for the function. The functions r(R), g(G) and b(B) are increased in an amount of red, green and blue, respectively, in proportion to the parameters R, G and B. When R=0, G=0 and B=0, c(C) is black. When R=255, G=255 and B=255, c(C) is white.
In the case of judging whether or not the object is a human, it is judged whether or not an area in the image data such as c(C) of each pixels in the vicinity of the center of the image corresponds to a specific color. It is judged that the object is a human when c(C) is a color of seasan in the present embodiment since the present embodiment is designed for a xanthoderm. Conditions under which c(C) be seasan are R≈250, G≈240 and B≈150. These values or areas specified by the values can be varied depending on the photographing conditions when so required. In addition, other colors may be set as c(C) instead of seasan depending on the race of human. Plural colors may be set as c(C) on the condition of logical OR so that plural races may be recognized as human.
In the case where the “auto” is selected in the detector appointing portion 4, the detector selecting portion 33S selects, based on the result obtained from the judgment by the object judging portion 32S, either one of the first corresponding point detector 34 which employs the gradient method or the second corresponding point detector 35 which employs the correlation method for the detection of corresponding points. For example, in the case where the object Q is a human, the first corresponding point detector 34 is selected; in the case where the object Q is not a human, the second corresponding point detector 35 is selected.
In the case where the “portrait” or the “landscape” is selected in the detector appointing portion 4, the object judging portion 32S does not perform the judgment processing. In this case, the detector selecting portion 33S operates in such a manner that the first corresponding point detector 34 is selected if the “portrait” is selected and the second corresponding point detector 35 is selected if the “landscape” is selected.
The first corresponding point detector 34 detects corresponding points of plural image data DT2 by the gradient method. The second corresponding point detector 35 detects corresponding points of plural image data DT2 by the correlation method. Descriptions on the detectors will be given later in this specification.
The three-dimensional data generator 36 generates three-dimensional data DT3 based on the detection result obtained by either one of the first corresponding point detector 34 or the second corresponding point detector 35. More specifically, the three-dimensional data DT3 are generated by reproducing three-dimensional image of the object based on the corresponding points of the plural image data DT2. The three-dimensional data memory 37 memorizes the three-dimensional data DT3.
The image display 38 displays a three-dimensional image on the display 11 based on the three-dimensional data DT3. A liquid crystal display or the like can be used as the display 11.
The detection of the corresponding points by the gradient method in the first corresponding point detector 34 will be described below.
FIG. 3 is an example of plural image IM2; FIG. 4 is a graph showing an expression of luminance gradient; FIG. 5 is an enlarged view of windows WA and WB; FIG. 6 shows noted pixels GA and GB as well as pixels in the vicinity of the noted pixels; and FIG. 7 illustrates an example of the method wherein a multiple resolution image is used.
Images IM2 a and IM2 b are images that are represented by each of the plural image data DT2 a and DT2 b. The image data DT2 a and DT2 b are data generated by separate image input portions 21. The image IM2 b represents a standard image and the image IM2 a represents a reference image. The images IM2 a and IM2 b respectively include the object image QA and QB of the object Q. The windows WA and WB are set in each of the images IM2 a and IM2 b. The window WB is a standard window set in the standard image, and the window WA is a reference window set in the reference image. Here, the windows WA and WB are at a position that includes right eye of the object.
Referring to FIG. 5, each of the windows WA and WB is comprised of 9×9 pixels.
The windows each may be comprised of 3×3 pixels, 5×5 pixels, 7×7 pixels or other numbers of pixels. The noted pixels GA and GB respectively are pixels on coordinates that are in common with the windows WA and WB. As shown in FIG. 6, pixels neighboring to the pixel GA in vertical and horizontal directions respectively are defined as GA1, GA2, GA3 and GA4. The same applies to the pixel GB, and the pixel GB is adjacent to pixels GB1, GB2, GB3 and GB4.
Purpose for detecting the corresponding points is to find on the reference image points corresponding to all of points on the object image QB of the object Q in the standard image. Vector connecting the corresponding points when the reference image and the standard image are superimposed is referred to as a “corresponding vector” in this specification, and the detection of corresponding points using the gradient method means the detection of corresponding vector (u, v) with respect to points on an image. The corresponding vector has a point of the standard image as its base point and has a point on the reference image as its end point.
More specifically, the corresponding vector represents difference between the images of the windows WB and WA and has relationship indicated by the following expression (2).
dE/dt=Exu+Eyv+Et=0  (2).
In the expression (2), Ex=δE/δx, Ey=δE/δy, u=δx/δt; v=δy/δt; Et=δE/δt; E represents luminance; x represents a coordinate in horizontal direction of an image; y represents a coordinate in vertical direction; and t represents time. Note that, since the images dealt with in the present embodiments are stereographical images, t corresponds to a distance between two cameras 12.
Since the object images QA and QB of FIG. 3 are prepared by photographing almost the same portion of the object Q, it is possible to assume that the luminance gradient of a point on a coordinate of the object image QA is the same as that of a point on the same coordinate of the object image QB.
Since the expression (2) contains two unknown letters u and v while containing only one equation, it can represent nothing more than that corresponding vectors are on a linear line as shown in FIG. 4. Therefore, another restraining condition is required between u and v in order to determine the corresponding vectors. Thus, there is made such an assumption that u and v are common to all the pixels in a certain area, for example, in an area of 3×3 pixels. Solution for u and v with respect to all the pixels are obtainable by solving, by the least square method, a simultaneous equation that is made by adding the restraining condition to the expression (2). It is possible to reduce influence of quantization error in imaging or noise of imaging element by using a wider area such as 5×5 pixels and 6×6 pixels.
If the corresponding vectors with respect to all the pixels are the same, there can be prepared 81 types of the expression (2). The corresponding vectors are obtainable by solving the expressions by using the least square method.
For example, Ex, Ey and Et of the expression (2) with respect to the pixels GA and GB can be represented by the following expressions (4) to (6).
Ex={([GA 1]−[GA 2])+([GB 1 ]−[GB 2])}/2  (4)
Ey={([GA 3]−[GA 4])+([GB 3]−[GB 4])}/2  (5)
Et=[GB]−[GA]  (6)
In above expressions, the reference characters such as [GA1] respectively represent luminance values of the pixels shown in FIG. 6. For example, [GA1] represents the luminance value of the pixel GA1 shown in FIG. 6.
Ex, Ey and Et with respect to other pixels are obtained in the same manner.
However, if the corresponding vectors are obtained by above method, difference of images of only a few pixels can be precisely measured. More specifically, in the case where the corresponding points are different to each other by more than a several pixels due to a relatively large parallax, it is impossible to obtain a precise corresponding point. This is a constraint imposed by a premise of the gradient method and, therefore, the reference image and the standard image must be similar to each other in order to obtain the corresponding point.
In the case where the corresponding points are distant from each other, the multi resolution image is utilized in order to detect the corresponding points precisely. The method using the multi resolution image it self is known in the art.
As shown in FIG. 7, the multiple resolution image FM comprises an original image F and a plurality of types of images F1, F2 and F3 obtained by so compressing the original image F as to be (½)2, (¼)2, (⅛)2 or the like that are low in resolution. An image that is relatively low in resolution typically has smaller number of pixels than that of an image that is high in resolution. Therefore, pyramid shape is formed by arranging an image that is the highest in resolution at the bottom and other images in the descending order of resolution. Thus, the multiple resolution image FM forms an image pyramid.
By the use of the multiple resolution image FM, two positions distant to each other between images IM2 a and IM2 b when the resolution is high, i.e., when there is a large number of pixels, approaches to each other when there is a small number of pixels. Accordingly, it is possible to widen an area wherein the corresponding points are precisely obtained by the gradient method as a result of transmitting the difference of images from a low resolution to a high resolution.
In the case of using the multiple resolution image as mentioned above, the corresponding vector is obtained with respect to the lowest resolution image to be set as an initial value. Then, another corresponding vector is obtained with respect to the second lowest resolution image. By repeating above processing, corresponding vectors high in resolution are obtainable if points on the images IM2 a and IM2 b that represents the same point on the object Q are largely different to each other.
The detection of corresponding points by using the second corresponding point detector 35 is described below.
FIG. 8 shows a concept of detection using an epipolar line EL. FIG. 9 illustrates a relationship between a correlation value and difference.
The correlation method is a method for detecting windows that are high in correlation and is useful for calculating correlation between windows of two images.
In the calculation of correlation, a window that is the lowest in the correlation value is detected first. The correlation value is obtainable by adding differences between corresponding points of the windows. In the case of stereographical view, it is geometrically obvious that the corresponding windows are on the epipolar line and, therefore, the detection is conducted solely with respect to the epipolar line.
In FIG. 8, the window Wa is moved in the image IM2 a from left upper end to right bottom end, and a window Wb corresponding the window Wa is detected from the image IM2 b at each of the positions. The window Wb in the image IM2 b is shifted by one pixel from the same coordinate as that of the image IM2 a. Here, the correlation value of the windows Wa and Wb is calculated to detect a position which is the lowest in the correlation value. The difference obtained hereby is then obtained with respect to all the points moved in the image IM2 a to form a map of differences. The differences correspond to parallax in the case of the stereographical view and thus called “parallax map”.
In FIG. 9, the difference is used to enter the horizontal axis and the correlation value is used to enter the vertical axis. In the graph plotted by solid line, the correlation value is the lowest when the difference is detected with respect to seven pixels. Accordingly, it is understood from the graph that a window that is seven pixels off from the noted window is the most correlative to the noted window. Parallax obtained by above calculation is seven pixels.
Thus, parallax is obtainable with respect to pixels in the correlation method. However, the parallax does not always correspond to number of pixels. In order to obtain parallax with respect to subpixels by the correlation method, the graph of solid line is interpolated. One of the methods for the interpolation is a method wherein four points close to the lowest correlation value are noted and a point of intersection of lines formed by connecting each two points of the four points is set as the lowest correlation value.
By using above interpolation method, the parallax corresponds to 7.3 pixels as shown by the broken line in FIG. 9.
Generation of three-dimensional data DT3 according to the first embodiment will be described below with reference to a flowchart.
FIG. 10 is the flowchart showing a flow of the generation of three-dimensional data DT3 according to the first embodiment.
As shown in FIG. 10, two sets of image data DT2 are obtained by the two image input portions 21 (#11). The camera 12 used for photographing the object Q is subjected to calibration in advance of the photographing in order to grasp camera parameters such as projection matrix and lens aberration. In the case of photographing a moving object, the object is shoot in succession by monocular vision using one camera.
Amounts of colors of the object Q are detected based on the obtained image data DT2 (#12) to judge whether or not the object is a human (#13).
If it is judged that the object Q is a human, a corresponding point of the image data DT2 a and DT2 b is detected by the first corresponding point detector 34, i.e. by the gradient method (#14). If it is not judged that the object Q is a human, the corresponding point of the image data DT2 a and DT2 b is detected by the second corresponding point detector 35, i.e. by the correlation method (#15).
Then, three-dimensional data DT3 are generated based on either of the detection results of the step 14 or the step 15 (#16).
As described above, it is possible to generate highly reliable three-dimensional data by judging types of the object by detecting its colors, followed by selecting a detector, or a detection method, for detecting a corresponding point based on the judgment result.
In addition, it is possible to judge whether or not the object is a human based on any colors. Further, if the color r(R) solely is extracted from pixels of the object Q. characteristics of eyes, nose, forehead or cheeks are represented by red color gradation. It is possible to judge whether or not the object Q is a human by the red color gradation. The judgment may be made based on other various information such as infrared rays that radiates from the object by other various methods.
(Second Embodiment)
In the first embodiment, the selection of detector for generating the three-dimensional data is performed by detecting colors of the object. In turn, the selection is performed by calculating luminance gradient in the second embodiment.
FIG. 11 is a block diagram showing a three-dimensional data generating device 1B according to the second embodiment of the present invention.
As shown in FIG. 11, the three-dimensional data generating apparatus 1B comprises two image input portions 21 and a processor 3B. The image input devices are the same as those of the first embodiment.
The processor 3B of the second embodiment comprises a luminance gradient calculator 32T and a detector selecting portion 33T unlike the processor 3A of the first embodiment. The two image data memories 31, the first corresponding detector 34, the second corresponding point detector 35, the three-dimensional data generator 36, the three-dimensional data memory 37 and the image display 38 are the same as those of the first embodiment. Operation processing is carried out in the luminance gradient calculator 32T only when the “auto” is appointed by the detector appointing portion 4, and the detector selecting portion 33T executes the selection based on the calculation results in the same manner as the first embodiment. In the case where the “portrait” or the “landscape” is appointed by the detector appointing portion 4, the first corresponding point detector 34 or the second corresponding point detector 35 is selected based on the appointment in the same manner as the first embodiment.
The luminance gradient calculator 32T calculates a luminance gradient of an object based on image data DT2. A luminance gradient Ex in the horizontal direction is obtainable by the expression (4), and a luminance gradient in the vertical direction Ey is obtainable by the expression (5). The luminance gradients Ex and Ey are calculated with respect to a several points of the image data DT2, and the luminance gradient of the object is calculated by, for example, obtaining a mean luminance gradient of the points. Two sets of image data DT2 used in the expressions (4) and (5) are obtained by photographing almost the same part and, therefore, it is possible to use only one of them for the calculation.
The detector selecting portion 33T selects either one of the first corresponding pint detector 34 or the second corresponding point detector 35 based on the judgment result obtained by the luminance gradient calculator 32T. The luminance gradient calculator 32T performs the selection in such a manner that the first corresponding point detector 34 is selected when the luminance gradient is lower than a predetermined value while the second corresponding point detector 35 is selected when the luminance gradient is higher than the predetermined value.
Processing for generating three-dimensional data according to the second embodiment will be described below with reference to a flowchart.
FIG. 12 is the flowchart showing the processing for generating three-dimensional data according to the second embodiment.
As shown in FIG. 12, two sets of image data are obtained by two image input portions 21 in the same manner as the first embodiment (#21).
A luminance gradient of an object is calculated based on the obtained image data (#22), and it is judged whether or not a result of the calculation is higher than a predetermined value (#23).
In the case where the calculation result is higher than the predetermined value, a corresponding point of the two sets of image data is detected by using the second detector—35, i.e. by the correlation method (#24). In the case where the calculation result is lower than the predetermined value, the corresponding point of the two sets of image data is detected by the first detector—34, i.e. by the gradient method (#25).
Then, in the same manner as the first embodiment, three-dimensional data is generated based on the detection result (#26).
As described above, it is possible to generate the highly reliable three-dimensional data by calculating the luminance gradient of the object and selecting the detector for the corresponding point based on the calculation result according to the second embodiment.
(Third Embodiment)
In the first and second embodiments, generation of three-dimensional data is performed by using either one of the detectors. In turn, in the third embodiment, both of the detectors are used for the detection of corresponding point and the generation of three-dimensional data is performed by using detection result of higher reliability.
FIG. 13 shows an appearance of a three-dimensional data generating apparatus 1C according to the third embodiment. FIG. 14 is a block diagram showing a structure of the three-dimensional data generating apparatus 1C according to the third embodiment.
As shown in FIGS. 13 and 14, the three-dimensional image generating apparatus 1C comprises three image input portions 21 and a processor 3C. The image input portions operates in the same manner as those of the first embodiment.
The processor 3C is comprised of the same components as those of the first embodiment; however, the processor 3C is provided with three image data memories 31, a reliability operation portion 32U, a reliability judging portion 33U and a three-dimensional data generator 36U. Functions of the first corresponding point detector 34, the second corresponding point detector 35, the three-dimensional data memory 37, the image display 38 and the like are realized in the same manner as the first embodiment.
The reliability judging portion 33U compares a detection result obtained by the first corresponding point detector with a detection result obtained by the second corresponding point detector to judge which detection result is higher in reliability.
The three-dimensional data generator 36U performs three-dimensional reproduction based on the detection result higher in reliability based on the judgment result of the reliability judging portion 33U.
A method of obtaining reliability of the detection results will be described below with reference to a flowchart and drawings.
FIG. 15 illustrates an example of the method of obtaining reliability of the detection results.
The image data DT2 e, DT2 f and DT2 g are respectively generated by different image input portions. The point Pt1 is an arbitrary point included in the image data DT2 e. The points Pt2 and Pt3 respectively are obtained by the first corresponding point detector and the second corresponding point detector as points on the image data DT2 f and DT2 g corresponding to the point Pt1. The point Pt12 is a point in a three-dimensional space obtained based on the points Ptl and Pt2. The point Pt3′ is a point obtained by projecting the point Pt12 on the image data DT2 g.
FIG. 16 is a flowchart showing a flow of processing for obtaining the reliability of detection results.
In FIG. 16, there are obtained the points Pt2 and Pt3 that are on the image data DT2 f and DT2 g, respectively, and correspond to the point Pt1 on the image data DT2 e (#91).
The point Pt12 on a three-dimensional coordinate is calculated based on the points Pt1 and Pt2 (#92). The point Pt12 is then projected onto the image data DT2 g to obtain Pt3′ (#93). The points Pt3 and Pt3′ are logically the same; however, in reality, the points Pt3 and Pt3′ are different from each other due to calibration error or characteristics of the camera 12.
The difference between the point Pt3 and the point Pt3′ is determined by comparison thereby to obtain reliability of the detection result (#94). Smaller the difference is, higher is the reliability.
Processing for generation of the three-dimensional data according to the third embodiment is described below with reference to a flowchart.
FIG. 17 is the flowchart showing a flow of the processing for generation of the three-dimensional data according to the third embodiment.
As shown in FIG. 17, three sets of image data are obtained by the three image input portions 21 (#31).
Detection of corresponding points are performed by the first corresponding point detector 34 and the second corresponding point detector 35 (#32), and reliability of the detection results are calculated (#33).
Based on the calculation results, it is judged which one of the detection result obtained by the first corresponding point detector 34 and the detection result obtained by the second corresponding point detector 35 is higher in the reliability (#34). Thus, the detection result higher in the reliability is selected (#35 and #36).
The three-dimensional data are generated based on the selected detection result (#37).
As described above, since the reliability of the detection result obtained by the first corresponding point detector 34 is compared with the reliability of the detection result obtained by the second corresponding point detector 35 and the three-dimensional data are generated based on the detection result higher in the reliability, it is possible to generate highly reliable three-dimensional data according to the third embodiment.
Further, the comparison of the reliability of the detection results is performed with respect to whole parts of the detection results in this embodiment; however, the comparison may be performed with dividing the detection results into partial areas.
For example, reliability of detection result of the first corresponding point detector 34 and reliability of detection result of the second corresponding point detector 35 are obtained with respect to a plurality of partial areas, and the reliabilities are compared to each other with respect to each of the partial areas to select the higher reliability with respect to each partial areas. Then, detection results higher in the reliability are so synthesized as to complete three-dimensional reproduction, thereby to generate the three-dimensional data.
Since the parts that are high in reliability are thus selected from the detection results obtained by both of the first corresponding point detector 34 and the second corresponding point detector 35, it is possible to generate three-dimensional data high in precision.
In the embodiments described above, parts corresponding to a background may be deleted from the image data in advance of the generation of three-dimensional data. Processing of the deletion of background may be carried out by various known methods. The deletion of background enables to eliminate redundant data and thus to reduce erroneous corresponding points.
In the embodiments described above, the image input portions 21 and the processor 3 are integrally comprised in the three-dimensional data generating apparatus 1, 1B and 1C; however, the image input portions and the processor 3 may be separately comprised in the three-dimensional data generating apparatus. FIG. 18 is a block diagram showing the modification of three-dimensional data generating apparatus. In this modification, a computer such as a personal computer is used for the processor 3. In FIG. 18, a multiple-lens camera 2 has a plurality of image input portions 21 and output portions 22A, 22B and 22C for outputting image signals obtained by the image input portions. Reference numeral 22A denotes a connection port for transferring the image signals to an external device via interface such as USB; reference numeral 22B denotes an output port for transferring the image signals to an external device via wireless communication such as infrared communication and reference numeral 22C denotes a driver for recording the image signals to a portable recording medium such as a memory card. In turn, a computer 3 is provided with input portions 30A, 30B and 30C that correspond respectively to the output portions 22A, 22B and 22C. The input portion 30A is a connection port for receiving signals from an external device via interface; the input portion 30B is an input port for receiving signals from an eternal device via wireless communication and the input portion 30C is a driver for reading data memorized in a portable recording medium. An image data memory 31 serves to receive the image signals input into the input portions and to store the same.
Thus, it is possible to improve configuration of the multiple-lens camera by reducing functions and devices required for multiple-lens camera as a result of performing burdensome processing such as the generation of three-dimensional data using a computer.
Whole or parts of structure of the three-dimensional data generating apparatus 1, 1B or 1C, number of components, contents of processings or orders of processings may be modified as required in the scope and spirits of the present invention.
According to the embodiments described above, it is possible to generate highly reliable three-dimensional irrelevant to the type of an object.
Further, according to the embodiments described above, it is possible to select a detector suitable for detecting a corresponding point of plural image data easily. Moreover, it is possible to select a highly reliable detection result of corresponding point to generate three-dimensional data of high precision.

Claims (15)

1. A three-dimensional data generating apparatus, comprising:
an obtaining portion for obtaining plural image data representing images obtained by photographing an object from different viewpoints, the plural image data each representing respective one of the images;
a first detector for detecting corresponding points of the images by a gradient method;
a second detector for detecting corresponding points of the images by a correlation method;
a selector for selecting either one of the first detector or the second detector; and
a generator for generating three-dimensional image of the object based on the corresponding points detected by the selected detector.
2. The three-dimensional data generating apparatus according to claim 1, further comprising:
image data generators each of which generates image data by photographing an object, wherein
the obtaining portion obtains the image data each generated by the respective image data generators.
3. The three-dimensional data generating apparatus according to claim 1, further comprising:
an input portion for obtaining an image data from outside of the apparatus, wherein the obtaining portion obtains the image data from the input portion.
4. The three-dimensional data generating apparatus according to claim 1, further comprising:
an object judging portion for judging a type of an object, wherein the selector selects either one of the first detector or the second detector based on the judgment result of the object judging portion.
5. The three-dimensional data generating apparatus according to claim 4, wherein
the object judging portion judges whether or not the object is a human; and
the selector selects either one of the first detector or the second detector in such a manner that the first detector is selected if it is judged that the object is a human and the second detector is selected if it is judged that the object is not a human.
6. The three-dimensional data generating apparatus according to claim 5, wherein
the object judging portion detects colors of an object and judges that the object is a human if a predetermined amount of a predetermined color is contained in the object.
7. The three-dimensional data generating apparatus according to claim 1, further comprising:
a luminance gradient calculation portion for calculating a luminance gradient of an object, wherein
the selector selects either one of the first detector or the second detector in such a manner that the first detector is selected if the calculated luminance is lower than a predetermined value and the second detector is selected if the calculated luminance gradient is higher than the predetermined value.
8. The three-dimensional data generating apparatus according to claim 1, further comprising:
an appointing portion for appointing either one of the detectors, wherein
the selector selects either one of the first detector or the second detector based on the appointment by the appointing portion.
9. A three-dimensional data generating apparatus, comprising:
an obtaining portion for obtaining plural image data representing images obtained by capturing an object from different viewpoints, the image data each representing respective one of the images;
a first detector for detecting corresponding points of the images by a gradient method;
a second detector for detecting corresponding points of the images by a correlation method;
a precision operation portion for operating a precision of the detection of the corresponding points carried out by the first detector based on the corresponding points and a precision of detection of the corresponding points carried out by the second detector based on the corresponding points;
a decision portion for deciding to use either one of the corresponding points detected by the first detector or the corresponding points detected by the second detector based on the precision obtained by the precision operation portion; and
a generator for generating a three-dimensional data of the object based on the corresponding points decided by the decision portion.
10. The three-dimensional data generating apparatus according to claim 9, further comprising:
a first image data generator, a second image data generator and a third image data generator, each of which generating image data by photographing an object, wherein
the precision operation portion operates the precision of the detection by obtaining a point corresponding to a reference point in the image data generated by the first image data generator in each of the image data generated by the second and the third image data generators and comparing one of the corresponding—points with the other.
11. A three-dimensional data generating method, comprising the steps of:
obtaining as plural image data plural images respectively obtained by photographing an object from different angles;
selecting either one of a gradient method or a correlation method;
generating three-dimensional data of the object based on corresponding points obtained by the selected method, the corresponding points being a pair of points indicating an identical part of the object in the plural images.
12. The three-dimensional data generating method according to claim 11, wherein
either one of the gradient method or the correlation method is selected based on the plural images.
13. The three-dimensional data generating method according to claim 12, wherein
either one of the gradient method or the correlation method is selected based on luminance of the plural images.
14. The three-dimensional data generating method according to claim 13, wherein
either one of the gradient method or the correlation method is selected based on color composition of the plural images.
15. The three-dimensional data generating apparatus according to claim 11, wherein
either one of the gradient method or the correlation method is selected based on a comparison of reliabilities of the corresponding points obtained by the gradient method with reliabilities of the corresponding points obtained by the correlation method.
US09/925,742 2000-08-11 2001-08-10 Method and apparatus for generating three-dimensional data of an object by selecting the method for detecting corresponding points that is suitable to the object Expired - Lifetime US6996266B2 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2000243860A JP4193342B2 (en) 2000-08-11 2000-08-11 3D data generator
JP2000-243860 2000-08-11

Publications (2)

Publication Number Publication Date
US20020018062A1 US20020018062A1 (en) 2002-02-14
US6996266B2 true US6996266B2 (en) 2006-02-07

Family

ID=18734649

Family Applications (1)

Application Number Title Priority Date Filing Date
US09/925,742 Expired - Lifetime US6996266B2 (en) 2000-08-11 2001-08-10 Method and apparatus for generating three-dimensional data of an object by selecting the method for detecting corresponding points that is suitable to the object

Country Status (2)

Country Link
US (1) US6996266B2 (en)
JP (1) JP4193342B2 (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100319796A1 (en) * 2009-06-23 2010-12-23 Whitaker Carl T Multi-Port Valve
US8743374B2 (en) * 2012-09-11 2014-06-03 Keyence Corporation Shape measuring device, shape measuring method, and shape measuring program

Families Citing this family (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040125114A1 (en) * 2002-12-31 2004-07-01 Hauke Schmidt Multiresolution image synthesis for navigation
US20050281992A1 (en) * 2004-06-21 2005-12-22 Chiu-Lang Lin Reflecting mark having anti-counterfeit arrangment
JP4862816B2 (en) * 2007-12-17 2012-01-25 コニカミノルタホールディングス株式会社 Image correspondence point search device, distance measuring device and image motion detection device using the same
JP2009168536A (en) * 2008-01-15 2009-07-30 Fujifilm Corp Three-dimensional shape measuring device and method, three-dimensional shape regenerating device and method, and program
CN101790103B (en) * 2009-01-22 2012-05-30 华为技术有限公司 Parallax calculation method and device
TWI435593B (en) 2010-12-30 2014-04-21 Altek Corp Method for capturing three-dimensional image
CN103959038B (en) * 2011-10-18 2016-05-04 卢米尼克斯股份有限公司 For the method and system of view data processing
US20130258129A1 (en) * 2012-03-28 2013-10-03 Qualcomm Incorporated Method and apparatus for managing orientation in devices with multiple imaging sensors
CA2848794C (en) * 2014-04-11 2016-05-24 Blackberry Limited Building a depth map using movement of one camera
CN105025193B (en) * 2014-04-29 2020-02-07 钰立微电子股份有限公司 Portable stereo scanner and method for generating stereo scanning result of corresponding object
JP2015045654A (en) * 2014-09-30 2015-03-12 洋彰 宮崎 Shape recognition machine
JP6984583B2 (en) * 2016-02-22 2021-12-22 ソニーグループ株式会社 Information processing equipment and information processing method
JP6843552B2 (en) * 2016-08-23 2021-03-17 キヤノン株式会社 Image processing equipment, image processing methods and programs.
JP2020159965A (en) * 2019-03-27 2020-10-01 東北電力株式会社 Three-dimensional measurement device, three-dimensional measurement system, and three-dimensional measurement method

Citations (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4062045A (en) * 1975-06-02 1977-12-06 The President Of Hokkaido University Three-dimensional television system
US5179441A (en) * 1991-12-18 1993-01-12 The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration Near real-time stereo vision system
JPH07103734A (en) 1993-10-01 1995-04-18 Sharp Corp Apparatus for searching stereocorrespondence
JPH07105382A (en) 1993-09-30 1995-04-21 Toshiba Corp Moving picture processing unit
JPH0944677A (en) 1995-08-03 1997-02-14 Canon Inc Correspondent point extracting device
US5606627A (en) * 1995-01-24 1997-02-25 Eotek Inc. Automated analytic stereo comparator
JP2000092454A (en) 1998-09-14 2000-03-31 Sony Corp Device and method for image information conversion
US6064775A (en) * 1996-12-27 2000-05-16 Fuji Xerox Co., Ltd. Image processing apparatus for enhancing texture and three dimensional effect
JP2000172741A (en) 1998-12-11 2000-06-23 Meidensha Corp Method for calibrating drawing image
JP2000339473A (en) 1999-05-27 2000-12-08 Inst Of Physical & Chemical Res Method for tracking feature point in two-dimensional moving image
US6181815B1 (en) * 1997-02-25 2001-01-30 Nec Corporation Subject image extraction device
US6215899B1 (en) * 1994-04-13 2001-04-10 Matsushita Electric Industrial Co., Ltd. Motion and disparity estimation method, image synthesis method, and apparatus for implementing same methods
US6249616B1 (en) * 1997-05-30 2001-06-19 Enroute, Inc Combining digital images based on three-dimensional relationships between source image data sets
US6269182B1 (en) * 1995-05-23 2001-07-31 Olympus Optical Co., Ltd. Color measuring system using multispectral image data
US6389169B1 (en) * 1998-06-08 2002-05-14 Lawrence W. Stark Intelligent systems and methods for processing image data based upon anticipated regions of visual interest
US6434265B1 (en) * 1998-09-25 2002-08-13 Apple Computers, Inc. Aligning rectilinear images in 3D through projective registration and calibration
US6519358B1 (en) * 1998-10-07 2003-02-11 Sony Corporation Parallax calculating apparatus, distance calculating apparatus, methods of the same, and information providing media
US6580810B1 (en) * 1999-02-26 2003-06-17 Cyberlink Corp. Method of image processing using three facial feature points in three-dimensional head motion tracking
US6580821B1 (en) * 2000-03-30 2003-06-17 Nec Corporation Method for computing the location and orientation of an object in three dimensional space
US6606406B1 (en) * 2000-05-04 2003-08-12 Microsoft Corporation System and method for progressive stereo matching of digital images
US6608927B2 (en) * 1994-03-31 2003-08-19 Canon Kabushiki Kaisha Color image processing method and apparatus utilizing the same
US6611622B1 (en) * 1999-11-23 2003-08-26 Microsoft Corporation Object recognition system and process for identifying people and objects in an image of a scene
US6621921B1 (en) * 1995-12-19 2003-09-16 Canon Kabushiki Kaisha Image processing apparatus
US6674902B1 (en) * 1998-07-31 2004-01-06 Canon Kabushiki Kaisha Image processing apparatus image processing method, and storage medium
US6788809B1 (en) * 2000-06-30 2004-09-07 Intel Corporation System and method for gesture recognition in three dimensions using stereo imaging and color vision

Patent Citations (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4062045A (en) * 1975-06-02 1977-12-06 The President Of Hokkaido University Three-dimensional television system
US5179441A (en) * 1991-12-18 1993-01-12 The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration Near real-time stereo vision system
JPH07105382A (en) 1993-09-30 1995-04-21 Toshiba Corp Moving picture processing unit
JPH07103734A (en) 1993-10-01 1995-04-18 Sharp Corp Apparatus for searching stereocorrespondence
US6608927B2 (en) * 1994-03-31 2003-08-19 Canon Kabushiki Kaisha Color image processing method and apparatus utilizing the same
US6215899B1 (en) * 1994-04-13 2001-04-10 Matsushita Electric Industrial Co., Ltd. Motion and disparity estimation method, image synthesis method, and apparatus for implementing same methods
US5606627A (en) * 1995-01-24 1997-02-25 Eotek Inc. Automated analytic stereo comparator
US6269182B1 (en) * 1995-05-23 2001-07-31 Olympus Optical Co., Ltd. Color measuring system using multispectral image data
JPH0944677A (en) 1995-08-03 1997-02-14 Canon Inc Correspondent point extracting device
US6621921B1 (en) * 1995-12-19 2003-09-16 Canon Kabushiki Kaisha Image processing apparatus
US6064775A (en) * 1996-12-27 2000-05-16 Fuji Xerox Co., Ltd. Image processing apparatus for enhancing texture and three dimensional effect
US6181815B1 (en) * 1997-02-25 2001-01-30 Nec Corporation Subject image extraction device
US6249616B1 (en) * 1997-05-30 2001-06-19 Enroute, Inc Combining digital images based on three-dimensional relationships between source image data sets
US6389169B1 (en) * 1998-06-08 2002-05-14 Lawrence W. Stark Intelligent systems and methods for processing image data based upon anticipated regions of visual interest
US6674902B1 (en) * 1998-07-31 2004-01-06 Canon Kabushiki Kaisha Image processing apparatus image processing method, and storage medium
JP2000092454A (en) 1998-09-14 2000-03-31 Sony Corp Device and method for image information conversion
US6434265B1 (en) * 1998-09-25 2002-08-13 Apple Computers, Inc. Aligning rectilinear images in 3D through projective registration and calibration
US6519358B1 (en) * 1998-10-07 2003-02-11 Sony Corporation Parallax calculating apparatus, distance calculating apparatus, methods of the same, and information providing media
JP2000172741A (en) 1998-12-11 2000-06-23 Meidensha Corp Method for calibrating drawing image
US6580810B1 (en) * 1999-02-26 2003-06-17 Cyberlink Corp. Method of image processing using three facial feature points in three-dimensional head motion tracking
JP2000339473A (en) 1999-05-27 2000-12-08 Inst Of Physical & Chemical Res Method for tracking feature point in two-dimensional moving image
US6611622B1 (en) * 1999-11-23 2003-08-26 Microsoft Corporation Object recognition system and process for identifying people and objects in an image of a scene
US6580821B1 (en) * 2000-03-30 2003-06-17 Nec Corporation Method for computing the location and orientation of an object in three dimensional space
US6606406B1 (en) * 2000-05-04 2003-08-12 Microsoft Corporation System and method for progressive stereo matching of digital images
US6788809B1 (en) * 2000-06-30 2004-09-07 Intel Corporation System and method for gesture recognition in three dimensions using stereo imaging and color vision

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Frédéric Devernay Olivier Faugeras, "Computing Differential Properties of 3-D Shapes from Stereoscopic Images without 3-D Models", Programme 4-Robotique, image et vision Projet Robotvis, Rapport de recherche n <SUP>o </SUP>2304, Jul. 1994. *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100319796A1 (en) * 2009-06-23 2010-12-23 Whitaker Carl T Multi-Port Valve
US8743374B2 (en) * 2012-09-11 2014-06-03 Keyence Corporation Shape measuring device, shape measuring method, and shape measuring program

Also Published As

Publication number Publication date
US20020018062A1 (en) 2002-02-14
JP2002056379A (en) 2002-02-20
JP4193342B2 (en) 2008-12-10

Similar Documents

Publication Publication Date Title
US6996266B2 (en) Method and apparatus for generating three-dimensional data of an object by selecting the method for detecting corresponding points that is suitable to the object
KR102483838B1 (en) Multi-Baseline Camera Array System Architecture for Depth Augmentation in VR/AR Applications
US10390005B2 (en) Generating images from light fields utilizing virtual viewpoints
EP3149931B1 (en) Image processing system, imaging apparatus, image processing method, and computer-readable storage medium
US7215364B2 (en) Digital imaging system using overlapping images to formulate a seamless composite image and implemented using either a digital imaging sensor array
CA2819956C (en) High accuracy camera modelling and calibration method
EP2779623B1 (en) Apparatus and method for multispectral imaging with parallax correction
CN102124749B (en) Stereoscopic image display apparatus
US20140002612A1 (en) Stereoscopic shooting device
US20130101263A1 (en) Image capture device, player, system, and image processing method
US20150334373A1 (en) Image generating apparatus, imaging apparatus, and image generating method
JPWO2019026287A1 (en) Imaging device and information processing method
CN105791793A (en) Image processing method and electronic device
JPH1062140A (en) Method and device for reconstruction of shape
KR20180000696A (en) A method and apparatus for creating a pair of stereoscopic images using least one lightfield camera
US20090245584A1 (en) Image processing apparatus, image processing method, and program
WO2019146419A1 (en) Image processing device, image processing method, and program
EP3407297B1 (en) Method and device for determining a characteristic of a display device
WO2023112971A1 (en) Three-dimensional model generation device, three-dimensional model generation method, and three-dimensional model generation program
US11928775B2 (en) Apparatus, system, method, and non-transitory medium which map two images onto a three-dimensional object to generate a virtual image
KR20110025083A (en) Apparatus and method for displaying 3d image in 3d image system
US11665330B2 (en) Dynamic-baseline imaging array with real-time spatial data capture and fusion
JP6696596B2 (en) Image processing system, imaging device, image processing method and program
EP3229470B1 (en) Efficient canvas view generation from intermediate views
KR101722969B1 (en) Apparatus for detecting moving object and method for controlling the same

Legal Events

Date Code Title Description
AS Assignment

Owner name: MINOLTA CO. LTD., JAPAN

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:HAMAGUCHI, TAKAYUKI;REEL/FRAME:012084/0346

Effective date: 20010726

FEPP Fee payment procedure

Free format text: PAYOR NUMBER ASSIGNED (ORIGINAL EVENT CODE: ASPN); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY

STCF Information on status: patent grant

Free format text: PATENTED CASE

FPAY Fee payment

Year of fee payment: 4

FPAY Fee payment

Year of fee payment: 8

FPAY Fee payment

Year of fee payment: 12